I’ve been asked this more than any other single question about linguistics. How are programming languages different from human languages?
The scope of this question is vast, and I think its many answers delve into the deepest inquiries about language. What is (and is not) a language? What does it mean to use language? How does the human mind use language to think?
It also hits on what might be the most hotly-debated philosophical point in contemporary linguistics: to what extent is the human mind like a machine?
Following this question leads to an answer that is itself about questions.
Programming languages are designed to instruct a computer how to perform a task. In this sense, every element of a programming language is an imperative (individual lines of code aren’t called “commands” for nothing, are they?).
But this distinction doesn’t get interesting until it’s dissected.
I’ve heard the argument that this imperative-only perspective isn’t strictly valid: you can, for example, command a computer to display the output to a computation. Or tell a computer to perform multi-step processes in which the computer itself may be forced to make decisions using various algorithms. Isn’t this the artificial analogue to asking a question?
I would say that it is not. A human could (theoretically) perform any computation that s/he asks of a computer. If I were to ask a computer and a machine to perform the same task or process and give both parties identical instructions, the resulting product would be identical between the two parties.
Of course, that’s assuming neither computer nor human makes any mistakes. And the human may take much longer to come up with an answer.
But the point remains: when a user “asks” the computer a question, the user could theoretically work out the answer for him or herself. The user essentially outsources his or her processing power to the computer.
This is crucially different from questions in natural human languages: although many questions (such as “what’s fifteen percent tip on this bill?”) may be equated to the questions we ask of our computers, questions such as “what are you doing here?” or “what was your childhood like?” or “when is the last time you felt happy?” do not have responses that the asker could answer independently.
In natural human language, questions can be attempts to resolve a state that the asker cannot discover. So language becomes the only conduit through which such information can be passed. The human asker does not (and cannot) outsource his or her processing power; instead, s/he relinquishes his or her ability to independently discover this unknown state.
I call these “language reliant questions.” In these sorts of questions, language is the only conduit through which the asker can discover an answer to his or her inquiry. These questions have no universal truth value to their asker. Only the person to whom the question is asked can know the real truth value of the response (and, if he/she chooses to invent an answer, even the respondent may not know the question’s truth value).
In the other sort of question, language is only one (and perhaps not the most reliable) of a variety of methods for discovering an answer. I call this flavor of inquiry “language independent questions.” Asking “what’s fifteen percent tip on this bill?”, for instance, may assume pragmatic factors including but not limited to:
In every case, however, the asker may find the answer without language given a method for the answer’s discovery.
Sometimes, the distinction between these questions is not syntactic but contextual. “Is it raining outside?” may be language independent when asked by a human capable of going outside - in this context, the asker is motivated by factor 1 above. But the question may be language reliant when asked by a prisoner in a dungeon.
I know about Bloom’s Taxonomy and a few other crossovers between linguistics and the philosophy of questions, but I do not know if there’s an agreed-upon method for distinguishing between these sorts of questions. I think the binary distinction between these questions is crucial when understanding not only how programming languages differ from natural languages but how contemporary computers differ from humans.
Natural human languages can support both language reliant and language independent questions. Programming languages, however, cannot support language independent questions because humans control the methodology that computers use to discover answers. Since the computer’s methodology is always discoverable, we cannot ask computers a language independent question.
It’s funny, then, that the only almost-exception that I know of to this rule relates to computational language modeling. Linguists often use neural networks to test theories for how humans do things like form plurals or conjugate verbs. Neural networks do, in essence, invent their own methods for discovering the correct form of words or sentences given a combination of parameters and yes-or-no feedback. But even these networks can always be opened and examined in a way that human brains cannot be. Yet.
This is where you could stop me. After all, if computers can present their methodology, how do they do it if not through programming languages? Doesn’t the computer’s language become the only conduit by which a human discovers the answer to a question?
Yes. But, unlike human language, the language through which computers share their methodology can be further broken up into imperatives — unambiguous commands that humans can follow to independently discover the answer to a given question.
Compare this process with a theoretical methodology for how a respondent discovered the answer to “when is the last time you felt happy?”
So two questions — “how are programming languages different from human languages?” and “to what extent is the mind like a machine?” — share a complex, interrelated answer that is itself concerned with questions.
Someday, an ambiguous language (and therefore a language more similar to natural human languages) may allow us to talk to computers in a fundamentally different way. Someday, we may be able to ask computers questions and be unable to discover how the computer came upon its answer.
For now, though, there’s one question that questions can answer.
When I say “I went to the bank,” I could mean that I was exchanging US dollars for Turkish lira. I could also mean that I went to catch frogs. Whether I mean money-bank or river-bank depends on context.
I’ve been asked a few times why words can mean more than one thing. It doesn’t make immediate sense, after all, why natural language would have homophones — doesn’t it make more sense to have a one-to-one ratio between meaning and words?
Actually, languages tend toward the least number of words possible. Fewer words make a language easier to acquire.
Of course, too few words would make a language too ambiguous. So, if a word has multiple senses, the correct sense must be easy to resolve from context.
A language will — at least theoretically — minimize both lexical size and ambiguity. In the constant evolution of natural languages, the number of homophones in any language is a continuously-shifting compromise between the number of words in a language and the need for a word to be immediately understood.
A look at natural human languages reveals that homophones are generally not easily confusable except in rare contexts. If a word becomes too easily confused — again, theoretically — speakers will replace it with another.
If you want a real-life example, think about what happens when a person (let’s say, someone named John) joins a tight-knit group of friends who already have a friend named John. After a while, the group assigns one (or both) Johns a nickname to easily distinguish the two individuals.
The Yoruba language’s itutu literally translates to “dead.” Robert Farris Thompson of Yale University, however, translates the term as “mystic coolness.”
This may be the biggest lexical disagreement in translational history. However, Thompson argues that itutu does not so much describe death as it does the transcendental calm of the peacefully deceased.
This concept behind the word itutu broke out of art to define the aesthetic of Yoruba culture. Today, Thompson argues, itutu defines the Western aesthetic as well.
During the 15th century, the concept of Itutu spread all over West Africa. Thompson writes, “control, stability, and composure under the African rubric of the cool seem to constitute elements of an all-embracing aesthetic attitude.”
Thompson thinks African slaves preserved (or, as some cultural theorists would say, encoded) the concept of Itutu in African-American culture. This “cultural memory” was “decoded” almost four hundred years later by the black jazz musicians of the 1940s.
Credit for assigning the term “cool” to this ancient concept is credited to saxophonist Lester Young. And the term has persevered better than any jazz-era phrase. “Hipster” has made a comeback in the past few years, but no English speaker hears any jazz-era term quite as much as s/he hears “cool.” (Compare with the frequency of “daddy-o”).
And, after American soldiers became lexically infected with the cool/Itutu cultural construct, they spread their contagion to Europe during World War II. And it exploded from there.
You can view cool as African-American’s biggest impact on Western culture. A construct that clicked between cultures. Whatever it is, the term has proven to be one of the biggest semantic headaches for linguists.
How do you define cool? Thompson’s etymology helps. But, like “pornography,” “cool” is something you can’t define– but you know it when you see it.
Malcom Gladwell took a stab at the issue in his 1997 New Yorker Piece, “Who decides what’s cool?” Gladwell attempted to distinguish rather than define “cool.”
- The act of discovering what’s cool is what causes cool to move on
- Cool cannot be manufactured, only observed
- [Cool] can only be observed by those who are themselves cool.
And the more cynical (Frontline’s production, The Merchants of Cool, for instance) think there’s a lot of money in cool. Corporations can create an artificial cycle of cultural relevance. This method allows companies to define tastes for their consumers rather than allowing consumers to define their tastes for companies
Either way, Kurosawa looks pretty cool in that picture.
